I've run Bismark to align a set of BS-Seq data. Some (not all) of the samples had low mapping efficiency (~20%). I then tried mapping R1 and R2 separately...